# Models for inflation prediction

I am looking for some statistical models that are accepted to best predict inflation rates. Anything that is more complicated than linear model will be appreciated. Or even recommend books that discuss inflation predictions.

I would also be interested in models that could relate inflation to GDP(or some form of feature extracted from gdp).

-
Haha, good luck with that. There is no broad consensus whatsoever. On one front we have people who believe GDP and inflation are orthogonal, on the other we have a lot of people with a lot of ideas on what causes inflation - everyone with his own. I'm afraid in its current state, this question is too broad. – FooBar Dec 8 '14 at 16:00
Ok what about the banks and all the interested entities on the marked. Surely they already use some models for prediction. – Zee Dec 8 '14 at 16:05
Right, they do large-scale VARs and NK-models. Basically, they throw in anything they have and hope to increase $R^2$. However, this is not a practice that's is really accepted in academia. But, as you say, central banks gotta do something... So if you rephrase the question as this and remove academic acceptance, I think it's answerable. – FooBar Dec 8 '14 at 16:08
See pp391-395 here: www-personal.umich.edu/~samoore/bit885f2011/hendry-alchemy.pdf (sorry, I couldn't resist). – Ubiquitous Dec 8 '14 at 19:51
VAR, S-VAR, VEC, VAR-X, and in the last time bayesian VAR – user157623 Dec 9 '14 at 2:23

## 1 Answer

One approach that you might find useful/interesting would be Bayesian VARs.

An example of a Bayesian VAR being used to predict inflation can be found here. In this paper, Tim Cogley, Sergui Morozov, and Tom Sargent motivate some methodology for Bayesian VARs and then compare their results to what the Bank of England uses in their "fan charts" (which they use to predict inflation). Cogley and Sargent have two other papers on inflation that are related to this paper, "Drifts and Volatilities" and "Evolving Post-World War II U.S. Inflation Dynamics." This series of papers is a good expose on some of the theory of inflation and how to statistically approach it.

I believe @FooBar is right that most banks are using large-scale VARs (similar to the aforementioned) and New Keynesian models. This isn't directly related to the question, but there is a literature on model averaging in which they can combine the forecasts of VARs and N-K models to provide a "more informed" (I use the phrase more informed here lightly but I do think the model averaging is helpful) prediction as to what will happen. An example of such a paper can be found here

-
Thanks @cc7768 this is definitely useful and interesting information. However, it is about methods for prediction and not models. Whereas my question is about some general model(ignoring constants of course). – Zee Dec 10 '14 at 19:44
@cc7768 Nice information ;) could you give an example of a paper with model averaging of VAR and N-K models? – An old man in the sea. Sep 19 at 18:47